The Bags of Visual Words model is a traditional computer vision technique that represents an image as a collection of independent visual features. It was a leading mechanical feature extractor for image classification prior to the era of automatic representation learning via neural networks.

Claude

Before the widespread adoption of deep learning around 2012, extracting useful representations from images relied heavily on manual feature engineering. Researchers designed specialized feature extractors, such as SIFT (Scale-Invariant Feature Transform), SURF (Speeded Up Robust Features), HOG (Histograms of Oriented Gradients), and bags of visual words. These mechanically calculated descriptors formed the foundation of traditional computer vision pipelines, serving as inputs to classic machine learning models. Modern Convolutional Neural Networks (CNNs) largely replaced this paradigm by demonstrating that complex, hierarchical features could be learned directly from the raw image data.

Manual Feature Engineering in Computer Vision

Dive into Deep Learning

The Scale-Invariant Feature Transform (SIFT) is a traditional, mechanically calculated feature extractor used in computer vision. Prior to the advent of modern convolutional neural networks, SIFT was prominently used to manually engineer visual features from images for object detection and matching.

Scale-Invariant Feature Transform (SIFT)

Speeded Up Robust Features (SURF) is a handcrafted feature extraction algorithm used in traditional computer vision. Alongside other mechanical feature extractors, SURF was a standard method for representing image data before hierarchical feature learning became dominant.

Speeded Up Robust Features (SURF)

Histograms of Oriented Gradients (HOG) is a traditional feature descriptor used for object detection in computer vision. It mechanically calculates visual representations based on gradient orientations, serving as a primary method for feature engineering before the widespread adoption of deep learning.

Learn Before

Related